CLASSLA-web-sl-2 Enriched: Slovenian Web Corpus with Educational Metadata

4.79 million Slovenian web documents enriched with metadata for educational value, domain classification, and web registers. The dataset, created by zID4si, includes a filtered subset optimized for language model pre-training. It was last updated on February 13, 2026.

Use Cases

• Continued pre-training of language models based on the high-quality, filtered subset.
• Analyzing Slovenian web register distribution based on the annotated metadata.
• Studying the educational value of Slovenian web content based on the provided scores.
• Domain-specific text analysis based on the domain classification metadata.

Strengths

• 4.79 million documents provide a substantial text corpus.
• Includes a filtered subset optimized for language model training.
• Annotated with metadata for educational value, domain, and web registers.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the filtered subset is unknown, which may limit suitability assessment.
• Data may reflect temporal and source bias inherent to the original web crawl.

Provenance

Source

Enriched version of the Slovenian web corpus CLASSLA-web.sl.2.0.

Collection Method

Likely gathered via web crawling and subsequently annotated.

Freshness

Last updated 2026-02-13 19:25:40; freshness should be verified.

Geography

Slovenia (inferred from language focus).